Gas leakage false alarm prevention method and detection system based on sound source localization and visual identification

Abstract

The invention provides a gas leakage false alarm prevention method based on sound source localization and visual identification, which comprises the following steps of: acquiring multi-channel sound detection data acquired by an acoustic array, and acquiring image data acquired by an image sensor arranged at the center of the acoustic array; processing the multi-channel sound detection data based on a sound source positioning method by taking the predetermined area as a scanning plane to obtain at least one sound source position I; acquiring an audio information characteristic value for each sound source position I, and judging whether the audio information characteristic value accords with a gas leakage characteristic or not; if not, determining that the sound source position I is a false alarm point; if yes, intercepting a specified area image in the image data based on the sound source position I and a preset pixel range, identifying whether an object type in the specified area image is a target type or not based on a visual identification algorithm, and if not, judging that the sound source position I is a false alarm point; and if the type is the target type, determining that the sound source position I is a leakage point.

Description

Technical Field

[0001] The present invention relates to the technical field of gas leakage detection, and in particular to a gas leakage detection method and system based on sound source positioning and visual recognition. Background Art

[0002] In industries such as petrochemicals, electric power, and the automotive industry, pipeline and equipment inspection and maintenance have always been important tasks. Regular inspections by personnel can promptly detect abnormal conditions such as gas leaks, abnormal discharges, or mechanical vibrations, which is of great significance for ensuring production and the personal and property safety of personnel.

[0003] Traditional inspection methods rely on manual inspections, where inspectors rely on listening, observing, and watching for bubbles in equipment and pipelines. This approach is highly dangerous, inefficient, and prone to missed detections. Research has shown that gas leaks, abnormal electrical discharges, and mechanical vibrations all generate sound, leading to the development of acoustic imaging for sound source localization. Acoustic imaging utilizes highly sensitive digital MEMS microphones and an array signal processing algorithm to generate a sound pressure level distribution on a plane. This sound is visualized as a color contour map, and the sound distribution of the object under test is visually displayed through photos or videos. This method effectively compensates for the human ear's inability to locate the direction of a sound source, providing a more efficient solution for industrial inspections. For example, CN117028870A provides a natural gas leak location audio-visual system based on ultrasonic arrays and images. After processing multi-channel sound detection data based on the sound source localization method to obtain sound source location information, the sound source location information is directly superimposed and displayed with the image. However, in actual applications, wind, rain, natural sounds, etc. in nature may generate ultrasonic waves that overlap with the gas leak frequency band, thereby affecting the accuracy of the gas leak detection system.

[0004] In order to improve accuracy, a method of combining acoustic arrays with infrared, laser and other technologies has been further proposed. This method can improve the accuracy of detection, but it also greatly increases the cost of the detection system and has poor economic efficiency.

[0005] In order to solve the above problems, people have been seeking an ideal technical solution. Summary of the Invention

[0006] The purpose of the present invention is to address the deficiencies of the existing technology and thus provide a gas leak prevention false alarm method and system based on sound source positioning and visual recognition. The camera of the acoustic imaging device itself is used to visually identify the sound source position, identify whether it is a pipe, valve, pressure gauge or other location where leakage may occur, eliminate interfering sound sources, and improve recognition accuracy while more accurately locating the leakage position without increasing the hardware cost of the equipment.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is: A first aspect provides a method for preventing false alarms of gas leaks based on sound source localization and visual recognition, comprising the following steps: Acquire multi-channel sound detection data collected by the acoustic array, and acquire image data collected by the image sensor set at the center of the acoustic array, Using a predetermined area as a scanning plane, processing the multi-channel sound detection data based on a beamforming sound source localization method to obtain at least one sound source position I; For each sound source position I, an audio information characteristic value is obtained to determine whether the audio information characteristic value meets the gas leakage characteristic; if not, the sound source position I is determined to be a false alarm point; if it meets, a designated area image is intercepted in the image data based on the sound source position I and a preset pixel range, and whether the object type in the designated area image is the target type is identified based on a visual recognition algorithm; if not, the sound source position I is determined to be a false alarm point; if it is the target type, the sound source position I is determined to be a leakage point; wherein, the target types include at least pipelines, pipeline interfaces, valves, and pressure gauges.

[0008] Furthermore, after the object type in the designated area image is identified as the target type based on the visual recognition algorithm, the designated area image is used as the scanning plane, and the multi-channel sound detection data is processed based on the beamforming sound source localization method to obtain the sound source position II, and the sound source position II is used as the leakage point.

[0009] Furthermore, after identifying the object type in the designated area image as the target type based on the visual recognition algorithm, the local spatial position corresponding to the designated area image is obtained through coordinate transformation, and the acoustic array is controlled to perform multi-channel sound collection on the local spatial position to obtain multi-channel sound detection data. The designated area image is used as the scanning plane, and the multi-channel sound detection data is processed based on the beamforming sound source localization method to obtain the sound source position II, and the sound source position II is used as the leakage point.

[0010] Furthermore, the predetermined area is used as a scanning plane, and the multi-channel sound detection data is processed based on a low-resolution beamforming sound source localization method to obtain at least one sound source position I; The designated area image is used as the scanning plane, and the local multi-channel sound detection data is processed based on the high-resolution beamforming sound source localization method to obtain the sound source position II, which is then used as the leakage point.

[0011] Furthermore, before obtaining the audio information characteristic value for each sound source position I, it is necessary to first determine whether the sound source position I is within the field of view of the image sensor. If so, obtain the image data corresponding to the multi-channel sound detection data; if not, discard the sound source position I and continue to determine the next sound source position I.

[0012] Furthermore, identifying whether the object type in the image of the designated area is the target type based on the visual recognition algorithm includes: Use the trained feature extraction model to extract feature vectors from the specified area image and search in the preset feature vector library of the target type; If a feature vector that is consistent with the extracted feature vector is found in the feature vector library, the type of object in the specified area image is determined to be the target type; Among them, if a feature vector consistent with the extracted feature vector is not found in the feature vector library, the type of object corresponding to the specified area image is manually determined. If the object type is the target type, the extracted feature vector is added to the feature vector library of the target type.

[0013] A second aspect of the present invention provides a gas leak detection system based on sound source localization and visual recognition, comprising: Acoustic array, including a multi-channel microphone array, used to collect multi-channel sound detection data and output; An image sensor is provided at the center of the acoustic array and is used to collect and output real-time video image data; A gas leakage detection device, configured to process multi-channel sound detection data and real-time video image data based on the gas leakage detection method proposed in the first aspect to obtain a gas leakage point; The display device is used to superimpose and display the location of the leakage point, the corresponding type of object and the video image data after the gas leakage point is determined.

[0014] Furthermore, it also includes a video compression transmission device, which is used to store the video image data output by the image sensor and compress the video image data and send it to the gas leakage detection device; The video compression and transmission device is further configured to, upon receiving an image acquisition request from the gas leakage detection device, intercept an image of a designated area from the stored video image data and directly transmit the image to the gas leakage detection device, so that the gas leakage detection device replaces the designated area in the initial video image data with the acquired image of the designated area; The image acquisition request is generated by the gas leakage detection device based on the sound source position II and a preset pixel range after the sound source position II is taken as the leakage point.

[0015] A third aspect of the present invention provides a computer storage medium storing a target tracking program. When the target tracking program is executed by the processor, the steps of the gas leakage detection method based on sound source localization and visual recognition as described in the first aspect are implemented.

[0016] A fourth aspect of the present invention provides a computer program product, comprising a computer program, characterized in that when the computer program is executed by a processor, the steps of the gas leakage detection method based on sound source localization and visual recognition as described in the first aspect are implemented.

[0017] Compared with the existing technology, the present invention has outstanding substantial features and significant progress. Specifically, after locating the sound source position information based on the acoustic array, the present invention conducts a preliminary screening based on whether the sound source information at the sound source position meets the gas leakage characteristics. After passing the preliminary screening, the sound source position is visually identified using image data collected by an integrated camera to determine whether it is a location where leakage may occur, such as a pipeline, valve, pressure gauge, etc., and further eliminates interfering sound sources for a second time. While improving the recognition accuracy, the leakage position can be located more accurately without increasing the equipment hardware cost.

[0018] Furthermore, a predetermined area is first used as a scanning plane, and rapid sound source positioning and identification is performed based on the beamforming sound source localization method. After the identified sound source characteristics meet the gas leakage characteristics, the designated area image is intercepted in the image data based on the sound source position I for visual recognition. If the recognition result is the characteristics of a preset pipe, valve, pressure gauge or other object, the designated area image is used as the scanning plane, and the sound source positioning is performed again based on the beamforming sound source localization method to obtain the accurate leakage location.

[0019] Furthermore, when the predetermined area is used as the scanning plane, rapid sound source localization and identification is performed based on the low-resolution beamforming sound source localization method; when the designated area image is used as the scanning plane, accurate sound source localization is performed based on the high-resolution beamforming sound source localization method; compared with the direct high-resolution beamforming sound source localization method on the image data area, the two-step method is more efficient and requires less computation.

[0020] Furthermore, after the sound source position II is taken as the leakage point, uncompressed image data of the designated area is obtained based on the sound source position II to replace the compressed image, thereby improving the observation and display effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 1 is a flow chart of a method for preventing false alarms of gas leakage according to embodiment 1 of the present invention.

[0022] Figure 2 Schematic diagram of the structure of the acoustic array and image sensor according to embodiment 1 of the present invention.

[0023] Figure 3 2 is a flow chart of a method for preventing false alarms of gas leakage according to embodiment 2 of the present invention.

[0024] Figure 4 This is another schematic diagram of the gas leakage false alarm prevention method according to Example 2 of the present invention.

[0025] Figure 5 4 is a schematic structural diagram of a gas leakage detection system based on sound source localization and visual recognition according to Example 4 of the present invention.

[0026] In the figure, 1. Acoustic array; 2. Image sensor; 3. Gas leak detection device; 4. Display device. DETAILED DESCRIPTION

[0027] The technical solution of the present invention is further described in detail below through specific implementation methods.

[0028] Example 1 This embodiment provides a method for preventing false alarms of gas leaks based on sound source positioning and visual recognition. Figure 1 As shown, the following steps are included: Acquire multi-channel sound detection data collected by the acoustic array 1, and acquire image data collected by the image sensor 2 set at the center of the acoustic array 1. Specifically, the structures of the acoustic array and the image sensor are as follows: Figure 2 shown.

[0029] The predetermined area is used as a scanning plane, and the multi-channel sound detection data is processed based on the beamforming sound source localization method to obtain at least one sound source position I.

[0030] Specifically, the predetermined area is related to the resolution of the image sensor. Generally, the resolution of the image sensor is, for example, 1280*960, that is, the resolution of the image sensor is used as the predetermined area.

[0031] Specifically, the location of a sound source is generally represented by coordinates.

[0032] For each sound source position I, the audio information characteristic value corresponding to the position is obtained, and it is determined whether the audio information characteristic value meets the gas leakage characteristics.

[0033] Furthermore, determining whether the audio information characteristic value meets the gas leakage characteristic specifically includes: Acquire an audio information feature value of the sound source position, where the audio information feature value includes at least one of spectrum centroid, spectrum bandwidth, spectrum contrast, spectrum flatness, short-time zero-crossing rate, and Mel-frequency cepstral coefficient; The audio information characteristic value is compared with the preset characteristic value. If they are consistent, it is determined that the audio information characteristic value meets the gas leakage characteristic; otherwise, it is determined that the audio information characteristic value does not meet the gas leakage characteristic.

[0034] When the audio information characteristic value does not meet the gas leakage characteristics, the sound source position I is determined to be a false alarm point; at this time, if there is more than one sound source position I, the next sound source position I is determined until all sound source positions I are determined; if there is only one sound source position I, the determination is stopped.

[0035] When the audio information characteristic value meets the gas leakage characteristic, a designated area image is captured in the image data based on the sound source position I and the preset pixel range, and the type of object in the designated area image is identified based on the visual recognition algorithm to determine whether it is the target type. If it is not the target type, the sound source position I is determined to be a false alarm point; if it is the target type, the sound source position I is determined to be a leakage point; wherein, the target types include at least pipes, pipe interfaces, valves, and pressure gauges.

[0036] Specifically, the image data is generally synchronized with the multi-channel sound detection data, that is, collected at the same time.

[0037] Therefore, before obtaining the audio information characteristic value for each sound source position I, it is also necessary to first determine whether the sound source position I is within the field of view of the image sensor 2. If it is, obtain the image data corresponding to the multi-channel sound detection data; if not, discard the sound source position I and continue to determine the next sound source position I.

[0038] Specifically, the field of view is used to represent the field of view of the image sensor 2 under at least a predetermined pitch angle and azimuth angle configuration. In one embodiment, this step can be performed by determining whether the coordinates of the sound source position are within the field of view.

[0039] Specifically, when the video captured by the image sensor 2 is obtained, it can be first cached locally in the electronic device. When needed, the video frames can be read according to a preset rule to obtain each frame of image data. For example, the video frame images can be read frame by frame in chronological order to determine the video frame image data corresponding to the multi-channel acoustic signal.

[0040] In one embodiment, identifying whether an object in an image of a designated area is a target type based on a visual recognition algorithm includes: Use the trained feature extraction model to extract feature vectors from the specified area image and search in the preset feature vector library of the target type; If a feature vector that is consistent with the extracted feature vector is found in the feature vector library, the type of object in the specified area image is determined to be the target type; Among them, if a feature vector consistent with the extracted feature vector is not found in the feature vector library, the type of object corresponding to the specified area image is manually determined. If the object type is the target type, the extracted feature vector is added to the feature vector library of the target type.

[0041] In specific implementation, an existing feature extraction network can be selected to pre-construct a sample set, which includes several sample images. The sample images are images containing all or part of the target object captured from the existing scene environment image according to the specified area size. The feature extraction network is trained using the sample set to obtain a trained feature extraction network, and a feature vector library of the target type is constructed based on the extracted feature vectors.

[0042] After locating the sound source position information based on the acoustic array 1, the present invention performs a preliminary screening based on whether the sound source information at the sound source position meets the gas leakage characteristics. After passing the preliminary screening, the image data collected by the integrated image sensor 2 is used to visually identify the sound source position to identify whether it is a location where a leak may occur, such as a pipeline, valve, pressure gauge, etc., and further eliminate interfering sound sources for a second time. While improving the recognition accuracy, the leakage position can be located more accurately without increasing the equipment hardware cost.

[0043] Fast response time. This embodiment uses a predetermined area as a scanning plane and processes multi-channel sound detection data based on a beamforming sound source localization method. The lower-resolution sound source localization algorithm reduces computational complexity and shortens calculation time. It can quickly identify the audio characteristics of the leak sound source and locate the sound source point, allowing for rapid detection of gas leaks and prompt response and alarm.

[0044] Example 2 The difference between this embodiment and embodiment 1 is that: Figure 3 and Figure 4 As shown, after the object type in the designated area image M is identified as the target type based on the visual recognition algorithm, the designated area image is used as the scanning plane, and the multi-channel sound detection data is processed based on the beamforming sound source localization method to obtain the sound source position II, and the sound source position II is used as the leakage point.

[0045] Furthermore, when the designated area image is used as the scanning plane, the local multi-channel sound detection data is processed based on the high-resolution beamforming sound source localization method to obtain the sound source position II, and the sound source position II is used as the leakage point.

[0046] For example, if the designated area is 21 pixels centered on the pixel corresponding to the leak point, and the designated area image is used as the scanning plane, the designated area size is 3*7. In this case, the designated area image is divided into 20*20 grids, and the 20*20 resolution beamforming sound source localization method is used to process the multi-channel sound detection data. Because 20*20 is much larger than 3*7, it is a high-resolution beamforming sound source localization method.

[0047] When the predetermined area is used as the scanning plane, the multi-channel sound detection data is processed based on the low-resolution beamforming sound source localization method to obtain at least one sound source position I.

[0048] For example, if the multi-channel sound detection data is 64 multi-channel sound detection data and the predetermined area is 1280*960, the predetermined area is divided into 64*48 grids, and the 64*48 resolution beamforming sound source localization method is selected to process the multi-channel sound detection data; since 64*48 is much lower than 1280*960, it is a low-resolution beamforming sound source localization method.

[0049] The method of this embodiment requires only 64*48 + 20*20 computations, combined with a visual recognition algorithm, to achieve precise positioning in an acoustic space with a resolution of 1280*960 pixels. Compared to directly performing high-resolution beamforming on a designated 1280*960 area for sound source localization, the complexity of this embodiment's method is reduced by 1-(64*48 + 20*20) / (1280*960) = 99.7%, significantly improving system efficiency.

[0050] This embodiment uses a two-step sound source localization algorithm. First, low-resolution sound source localization is performed to determine the sound source position and eliminate interference sources. Then, a high-resolution sound source localization algorithm is performed within a specified area. This can improve the sound source localization resolution, shorten the calculation time, and greatly reduce the computational complexity of high-resolution acoustic imaging.

[0051] Furthermore, this embodiment only needs to collect multi-channel sound detection data once, and adopts low-resolution beamforming algorithm and high-resolution beamforming algorithm to locate the sound source based on different scanning planes in different actual situations. The object feature judgment of the sound source positioning point is combined with visual recognition to judge whether the object feature within a specific range around the sound source positioning point may cause gas leakage. It can realize the intelligent elimination of irrelevant interference sources, quickly identify false alarms and automatically eliminate them, improve detection accuracy and system operation efficiency, and can improve gas leak detection accuracy and resolution without increasing the cost of existing hardware, and has high execution efficiency and good economy.

[0052] Example 3 The difference between this embodiment and embodiment 2 is that: after the object type in the designated area image is identified as the target type based on the visual recognition algorithm, the local spatial position corresponding to the designated area image is obtained through coordinate transformation, the acoustic array is controlled to perform multi-channel sound collection on the local spatial position to obtain multi-channel sound detection data, the designated area image is used as the scanning plane, the multi-channel sound detection data is processed based on the beamforming sound source localization method, the sound source position II is obtained, and the sound source position II is used as the leakage point.

[0053] This embodiment reduces the amount of multi-channel sound by controlling the acoustic array to collect multi-channel sound at a local spatial position, thereby increasing the speed of processing the multi-channel sound detection data based on the beamforming sound source localization method when the designated area image is used as the scanning plane in the later stage.

[0054] Example 4 This embodiment provides a gas leak detection system based on sound source localization and visual recognition, such as Figure 4 Shown, including: The acoustic array 1 includes a multi-channel microphone array for collecting and outputting multi-channel sound detection data. Specifically, the array layout of the microphone array is not limited to a rectangular array, a central array, a multi-arm spiral array, an irregular array, etc. Different arrangement forms can be suitable for different application scenarios. This embodiment is set up using a multi-arm spiral array as an example.

[0055] The image sensor 2 is arranged at the center of the acoustic array 1 and is used to collect real-time video image data and output it; preferably, the image sensor is a camera.

[0056] A gas leakage detection device 3, configured to process multi-channel sound detection data and real-time video image data based on the gas leakage detection method described in Example 1, Example 2, or Example 3 to obtain a gas leakage point; The display device 4 is used to superimpose and display the position of the leakage point, the corresponding type of object and the video image data after the gas leakage point is determined.

[0057] After locating the sound source position information based on the acoustic array, the present invention conducts a preliminary screening based on whether the sound source information at the sound source position meets the gas leakage characteristics. After passing the preliminary screening, the sound source position is visually identified using image data collected by an integrated camera to identify whether it is a location where a leak may occur, such as a pipeline, valve, or pressure gauge. Interference sound sources are further eliminated for a second time, which improves the recognition accuracy and can more accurately locate the leakage position without increasing the equipment hardware cost.

[0058] Example 5 The difference between this embodiment and embodiment 4 is that: Figure 5As shown, it also includes a video compression transmission device, which is used to store the video image data output by the image sensor 2 and compress the video image data and send it to the gas leakage detection device 3; The video compression and transmission device is further configured to, upon receiving an image acquisition request from the gas leakage detection device 3, intercept an image of a designated area from the stored video image data and directly transmit the image to the gas leakage detection device, so that the gas leakage detection device 3 replaces the designated area in the initial video image data with the acquired image of the designated area; The image acquisition request is generated by the gas leakage detection device 3 based on the sound source position II and a preset pixel range after the sound source position II is taken as the leakage point.

[0059] Preferably, the image sensor 2 is an ultra-high-definition camera.

[0060] In this embodiment, the video data collected by the image sensor 2 is compressed and then transmitted, which can reduce the transmission volume. After the leak location is determined, an uncompressed high-definition image of a specific area around the leak location is further obtained for display. Compared with directly displaying all images in high definition, this embodiment can greatly reduce the transmission volume without affecting the subsequent viewing effect of the human eye.

[0061] Example 6 This embodiment provides a computer storage medium storing a target tracking program. When the target tracking program is executed by the processor, the steps of the gas leak detection method based on sound source localization and visual recognition as described in Example 1, Example 2, or Example 3 are implemented.

[0062] Example 7 This embodiment provides a computer program product, including a computer program, characterized in that when the computer program is executed by a processor, it implements the steps of the gas leakage detection method based on sound source localization and visual recognition as described in Example 1, Example 2, or Example 3.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention. They should all be included in the scope of the technical solution for protection of the present invention.

Claims

1. A method for preventing false alarms of gas leaks based on sound source localization and visual recognition, characterized in that: The following steps are involved: Acquiring multi-channel sound detection data collected by the acoustic array, and acquiring image data collected by an image sensor disposed at the center of the acoustic array; Taking a predetermined area as a scanning plane, processing the multi-channel sound detection data based on a beamforming sound source localization method to obtain at least one sound source position I; For each sound source position I, an audio information characteristic value is obtained to determine whether the audio information characteristic value conforms to the gas leakage characteristic; if not, the sound source position I is determined to be a false alarm point; if it is consistent, a designated area image is intercepted in the image data based on the sound source position I and a preset pixel range, and based on a visual recognition algorithm, it is determined whether the object type in the designated area image is a target type; if not, the sound source position I is determined to be a false alarm point; if it is a target type, the sound source position I is determined to be a leakage point; wherein the target types include at least pipelines, pipeline interfaces, valves, and pressure gauges.

2. A method for preventing false alarms of gas leakage based on sound source localization and visual recognition according to claim 1, characterized in that: After identifying the object type in the designated area image as the target type based on the visual recognition algorithm, the designated area image is used as the scanning plane, and the multi-channel sound detection data is processed based on the beamforming sound source localization method to obtain the sound source position II, and the sound source position II is used as the leakage point.

3. A method for preventing false alarms of gas leakage based on sound source localization and visual recognition according to claim 1, characterized in that: After identifying the object type in the designated area image as the target type based on the visual recognition algorithm, the local spatial position corresponding to the designated area image is obtained through coordinate transformation, and the acoustic array is controlled to perform multi-channel sound collection on the local spatial position to obtain multi-channel sound detection data. The designated area image is used as the scanning plane, and the multi-channel sound detection data is processed based on the beamforming sound source localization method to obtain the sound source position II, and the sound source position II is used as the leakage point.

4. A method for preventing false alarms of gas leakage based on sound source localization and visual recognition according to claim 2 or 3, characterized in that: Taking a predetermined area as a scanning plane, processing the multi-channel sound detection data based on a low-resolution beamforming sound source localization method to obtain at least one sound source position I; The designated area image is used as the scanning plane, and the local multi-channel sound detection data is processed based on the high-resolution beamforming sound source localization method to obtain the sound source position II, which is then used as the leakage point.

5. A method for preventing false alarms of gas leakage based on sound source localization and visual recognition according to claim 1, 2 or 3, characterized in that: Before obtaining the audio information characteristic value for each sound source position I, it is also necessary to determine whether the sound source position I is within the field of view of the image sensor. If so, obtain the image data corresponding to the multi-channel sound detection data; if not, discard the sound source position I and continue to determine the next sound source position I.

6. A method for preventing false alarms of gas leakage based on sound source localization and visual recognition according to claim 1, 2 or 3, characterized in that: Identifying whether the object type in the image of the specified area is the target type based on the visual recognition algorithm includes: Use the trained feature extraction model to extract feature vectors from the specified area image and search in the preset feature vector library of the target type; If a feature vector that is consistent with the extracted feature vector is found in the feature vector library, the type of object in the image of the specified area is determined to be the target type; If a feature vector consistent with the extracted feature vector is not found in the feature vector library, the type of object corresponding to the designated area image is manually determined; if the object type is the target type, the extracted feature vector is added to the feature vector library of the target type.

7. A gas leak detection system based on sound source localization and visual recognition, characterized in that: include: Acoustic array, including a multi-channel microphone array, for collecting and outputting multi-channel sound detection data; An image sensor, disposed at the center of the acoustic array, for collecting and outputting real-time video image data; A gas leakage detection device, used for processing multi-channel sound detection data and real-time video image data based on the gas leakage detection method according to any one of claims 1 to 6 to obtain a gas leakage point; The display device is used to superimpose and display the location of the leakage point, the corresponding type of object and the video image data after the gas leakage point is determined.

8. A gas leak detection system based on sound source localization and visual recognition according to claim 7, characterized in that: It also includes a video compression transmission device, which is used to store the video image data output by the image sensor and compress the video image data and send it to the gas leakage detection device; The video compression transmission device is further used to intercept the designated area image in the stored video image data and directly send it to the gas leakage detection device after receiving the image acquisition request sent by the gas leakage detection device, so that the gas leakage detection device replaces the designated area in the initial video image data with the acquired designated area image; The image acquisition request is generated by the gas leakage detection device based on the sound source position II and a preset pixel range after taking the sound source position II as the leakage point.

9. A computer storage medium, characterized in that The computer storage medium stores a target tracking program, and when the target tracking program is executed by the processor, the steps of the gas leakage false alarm prevention method based on sound source positioning and visual recognition as described in any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the gas leakage false alarm prevention method based on sound source positioning and visual recognition described in any one of claims 1 to 6 are implemented.

Images